]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/fs/block_dev.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | * Copyright (C) 2001 Andrea Arcangeli <andrea@suse.de> SuSE | |
6 | */ | |
7 | ||
1da177e4 LT |
8 | #include <linux/init.h> |
9 | #include <linux/mm.h> | |
10 | #include <linux/fcntl.h> | |
11 | #include <linux/slab.h> | |
12 | #include <linux/kmod.h> | |
13 | #include <linux/major.h> | |
7db9cfd3 | 14 | #include <linux/device_cgroup.h> |
1da177e4 LT |
15 | #include <linux/highmem.h> |
16 | #include <linux/blkdev.h> | |
66114cad | 17 | #include <linux/backing-dev.h> |
1da177e4 LT |
18 | #include <linux/module.h> |
19 | #include <linux/blkpg.h> | |
b502bd11 | 20 | #include <linux/magic.h> |
1da177e4 | 21 | #include <linux/buffer_head.h> |
ff01bb48 | 22 | #include <linux/swap.h> |
585d3bc0 | 23 | #include <linux/pagevec.h> |
811d736f | 24 | #include <linux/writeback.h> |
1da177e4 LT |
25 | #include <linux/mpage.h> |
26 | #include <linux/mount.h> | |
27 | #include <linux/uio.h> | |
28 | #include <linux/namei.h> | |
1368c4f2 | 29 | #include <linux/log2.h> |
ff01bb48 | 30 | #include <linux/cleancache.h> |
c94c2acf | 31 | #include <linux/dax.h> |
acc93d30 | 32 | #include <linux/badblocks.h> |
189ce2b9 | 33 | #include <linux/task_io_accounting_ops.h> |
25f4c414 | 34 | #include <linux/falloc.h> |
1da177e4 | 35 | #include <asm/uaccess.h> |
07f3f05c | 36 | #include "internal.h" |
1da177e4 LT |
37 | |
38 | struct bdev_inode { | |
39 | struct block_device bdev; | |
40 | struct inode vfs_inode; | |
41 | }; | |
42 | ||
4c54ac62 AB |
43 | static const struct address_space_operations def_blk_aops; |
44 | ||
1da177e4 LT |
45 | static inline struct bdev_inode *BDEV_I(struct inode *inode) |
46 | { | |
47 | return container_of(inode, struct bdev_inode, vfs_inode); | |
48 | } | |
49 | ||
ff5053f6 | 50 | struct block_device *I_BDEV(struct inode *inode) |
1da177e4 LT |
51 | { |
52 | return &BDEV_I(inode)->bdev; | |
53 | } | |
1da177e4 LT |
54 | EXPORT_SYMBOL(I_BDEV); |
55 | ||
2af3a815 TK |
56 | void __vfs_msg(struct super_block *sb, const char *prefix, const char *fmt, ...) |
57 | { | |
58 | struct va_format vaf; | |
59 | va_list args; | |
60 | ||
61 | va_start(args, fmt); | |
62 | vaf.fmt = fmt; | |
63 | vaf.va = &args; | |
64 | printk_ratelimited("%sVFS (%s): %pV\n", prefix, sb->s_id, &vaf); | |
65 | va_end(args); | |
66 | } | |
67 | ||
dbd3ca50 | 68 | static void bdev_write_inode(struct block_device *bdev) |
564f00f6 | 69 | { |
dbd3ca50 VG |
70 | struct inode *inode = bdev->bd_inode; |
71 | int ret; | |
72 | ||
564f00f6 CH |
73 | spin_lock(&inode->i_lock); |
74 | while (inode->i_state & I_DIRTY) { | |
75 | spin_unlock(&inode->i_lock); | |
dbd3ca50 VG |
76 | ret = write_inode_now(inode, true); |
77 | if (ret) { | |
78 | char name[BDEVNAME_SIZE]; | |
79 | pr_warn_ratelimited("VFS: Dirty inode writeback failed " | |
80 | "for block device %s (err=%d).\n", | |
81 | bdevname(bdev, name), ret); | |
82 | } | |
564f00f6 CH |
83 | spin_lock(&inode->i_lock); |
84 | } | |
85 | spin_unlock(&inode->i_lock); | |
86 | } | |
87 | ||
f9a14399 | 88 | /* Kill _all_ buffers and pagecache , dirty or not.. */ |
ff01bb48 | 89 | void kill_bdev(struct block_device *bdev) |
1da177e4 | 90 | { |
ff01bb48 AV |
91 | struct address_space *mapping = bdev->bd_inode->i_mapping; |
92 | ||
f9fe48be | 93 | if (mapping->nrpages == 0 && mapping->nrexceptional == 0) |
f9a14399 | 94 | return; |
ff01bb48 | 95 | |
f9a14399 | 96 | invalidate_bh_lrus(); |
ff01bb48 | 97 | truncate_inode_pages(mapping, 0); |
1da177e4 | 98 | } |
ff01bb48 AV |
99 | EXPORT_SYMBOL(kill_bdev); |
100 | ||
101 | /* Invalidate clean unused buffers and pagecache. */ | |
102 | void invalidate_bdev(struct block_device *bdev) | |
103 | { | |
104 | struct address_space *mapping = bdev->bd_inode->i_mapping; | |
105 | ||
106 | if (mapping->nrpages == 0) | |
107 | return; | |
108 | ||
109 | invalidate_bh_lrus(); | |
110 | lru_add_drain_all(); /* make sure all lru add caches are flushed */ | |
111 | invalidate_mapping_pages(mapping, 0, -1); | |
112 | /* 99% of the time, we don't need to flush the cleancache on the bdev. | |
113 | * But, for the strange corners, lets be cautious | |
114 | */ | |
3167760f | 115 | cleancache_invalidate_inode(mapping); |
ff01bb48 AV |
116 | } |
117 | EXPORT_SYMBOL(invalidate_bdev); | |
1da177e4 LT |
118 | |
119 | int set_blocksize(struct block_device *bdev, int size) | |
120 | { | |
121 | /* Size must be a power of two, and between 512 and PAGE_SIZE */ | |
1368c4f2 | 122 | if (size > PAGE_SIZE || size < 512 || !is_power_of_2(size)) |
1da177e4 LT |
123 | return -EINVAL; |
124 | ||
125 | /* Size cannot be smaller than the size supported by the device */ | |
e1defc4f | 126 | if (size < bdev_logical_block_size(bdev)) |
1da177e4 LT |
127 | return -EINVAL; |
128 | ||
129 | /* Don't change the size if it is same as current */ | |
130 | if (bdev->bd_block_size != size) { | |
131 | sync_blockdev(bdev); | |
132 | bdev->bd_block_size = size; | |
133 | bdev->bd_inode->i_blkbits = blksize_bits(size); | |
134 | kill_bdev(bdev); | |
135 | } | |
136 | return 0; | |
137 | } | |
138 | ||
139 | EXPORT_SYMBOL(set_blocksize); | |
140 | ||
141 | int sb_set_blocksize(struct super_block *sb, int size) | |
142 | { | |
1da177e4 LT |
143 | if (set_blocksize(sb->s_bdev, size)) |
144 | return 0; | |
145 | /* If we get here, we know size is power of two | |
146 | * and it's value is between 512 and PAGE_SIZE */ | |
147 | sb->s_blocksize = size; | |
38885bd4 | 148 | sb->s_blocksize_bits = blksize_bits(size); |
1da177e4 LT |
149 | return sb->s_blocksize; |
150 | } | |
151 | ||
152 | EXPORT_SYMBOL(sb_set_blocksize); | |
153 | ||
154 | int sb_min_blocksize(struct super_block *sb, int size) | |
155 | { | |
e1defc4f | 156 | int minsize = bdev_logical_block_size(sb->s_bdev); |
1da177e4 LT |
157 | if (size < minsize) |
158 | size = minsize; | |
159 | return sb_set_blocksize(sb, size); | |
160 | } | |
161 | ||
162 | EXPORT_SYMBOL(sb_min_blocksize); | |
163 | ||
164 | static int | |
165 | blkdev_get_block(struct inode *inode, sector_t iblock, | |
166 | struct buffer_head *bh, int create) | |
167 | { | |
1da177e4 LT |
168 | bh->b_bdev = I_BDEV(inode); |
169 | bh->b_blocknr = iblock; | |
170 | set_buffer_mapped(bh); | |
171 | return 0; | |
172 | } | |
173 | ||
4ebb16ca DW |
174 | static struct inode *bdev_file_inode(struct file *file) |
175 | { | |
176 | return file->f_mapping->host; | |
177 | } | |
178 | ||
189ce2b9 CH |
179 | #define DIO_INLINE_BIO_VECS 4 |
180 | ||
181 | static void blkdev_bio_end_io_simple(struct bio *bio) | |
182 | { | |
183 | struct task_struct *waiter = bio->bi_private; | |
184 | ||
185 | WRITE_ONCE(bio->bi_private, NULL); | |
186 | wake_up_process(waiter); | |
187 | } | |
188 | ||
189 | static ssize_t | |
190 | __blkdev_direct_IO_simple(struct kiocb *iocb, struct iov_iter *iter, | |
191 | int nr_pages) | |
192 | { | |
193 | struct file *file = iocb->ki_filp; | |
194 | struct block_device *bdev = I_BDEV(bdev_file_inode(file)); | |
195 | unsigned blkbits = blksize_bits(bdev_logical_block_size(bdev)); | |
196 | struct bio_vec inline_vecs[DIO_INLINE_BIO_VECS], *bvec; | |
197 | loff_t pos = iocb->ki_pos; | |
198 | bool should_dirty = false; | |
199 | struct bio bio; | |
200 | ssize_t ret; | |
201 | blk_qc_t qc; | |
202 | int i; | |
203 | ||
204 | if ((pos | iov_iter_alignment(iter)) & ((1 << blkbits) - 1)) | |
205 | return -EINVAL; | |
206 | ||
207 | bio_init(&bio); | |
208 | bio.bi_max_vecs = nr_pages; | |
209 | bio.bi_io_vec = inline_vecs; | |
210 | bio.bi_bdev = bdev; | |
211 | bio.bi_iter.bi_sector = pos >> blkbits; | |
212 | bio.bi_private = current; | |
213 | bio.bi_end_io = blkdev_bio_end_io_simple; | |
214 | ||
215 | ret = bio_iov_iter_get_pages(&bio, iter); | |
216 | if (unlikely(ret)) | |
217 | return ret; | |
218 | ret = bio.bi_iter.bi_size; | |
219 | ||
220 | if (iov_iter_rw(iter) == READ) { | |
221 | bio_set_op_attrs(&bio, REQ_OP_READ, 0); | |
222 | if (iter_is_iovec(iter)) | |
223 | should_dirty = true; | |
224 | } else { | |
225 | bio_set_op_attrs(&bio, REQ_OP_WRITE, REQ_SYNC | REQ_IDLE); | |
226 | task_io_account_write(ret); | |
227 | } | |
228 | ||
229 | qc = submit_bio(&bio); | |
230 | for (;;) { | |
231 | set_current_state(TASK_UNINTERRUPTIBLE); | |
232 | if (!READ_ONCE(bio.bi_private)) | |
233 | break; | |
234 | if (!(iocb->ki_flags & IOCB_HIPRI) || | |
235 | !blk_mq_poll(bdev_get_queue(bdev), qc)) | |
236 | io_schedule(); | |
237 | } | |
238 | __set_current_state(TASK_RUNNING); | |
239 | ||
240 | bio_for_each_segment_all(bvec, &bio, i) { | |
241 | if (should_dirty && !PageCompound(bvec->bv_page)) | |
242 | set_page_dirty_lock(bvec->bv_page); | |
243 | put_page(bvec->bv_page); | |
244 | } | |
245 | ||
246 | if (unlikely(bio.bi_error)) | |
247 | return bio.bi_error; | |
248 | iocb->ki_pos += ret; | |
249 | return ret; | |
250 | } | |
251 | ||
b2e895db | 252 | static ssize_t |
c8b8e32d | 253 | blkdev_direct_IO(struct kiocb *iocb, struct iov_iter *iter) |
b2e895db AM |
254 | { |
255 | struct file *file = iocb->ki_filp; | |
4ebb16ca | 256 | struct inode *inode = bdev_file_inode(file); |
189ce2b9 | 257 | int nr_pages; |
b2e895db | 258 | |
189ce2b9 CH |
259 | nr_pages = iov_iter_npages(iter, BIO_MAX_PAGES); |
260 | if (!nr_pages) | |
261 | return 0; | |
262 | if (is_sync_kiocb(iocb) && nr_pages <= DIO_INLINE_BIO_VECS) | |
263 | return __blkdev_direct_IO_simple(iocb, iter, nr_pages); | |
c8b8e32d | 264 | return __blockdev_direct_IO(iocb, inode, I_BDEV(inode), iter, |
fe0f07d0 JA |
265 | blkdev_get_block, NULL, NULL, |
266 | DIO_SKIP_DIO_COUNT); | |
b2e895db AM |
267 | } |
268 | ||
5cee5815 JK |
269 | int __sync_blockdev(struct block_device *bdev, int wait) |
270 | { | |
271 | if (!bdev) | |
272 | return 0; | |
273 | if (!wait) | |
274 | return filemap_flush(bdev->bd_inode->i_mapping); | |
275 | return filemap_write_and_wait(bdev->bd_inode->i_mapping); | |
276 | } | |
277 | ||
585d3bc0 NP |
278 | /* |
279 | * Write out and wait upon all the dirty data associated with a block | |
280 | * device via its mapping. Does not take the superblock lock. | |
281 | */ | |
282 | int sync_blockdev(struct block_device *bdev) | |
283 | { | |
5cee5815 | 284 | return __sync_blockdev(bdev, 1); |
585d3bc0 NP |
285 | } |
286 | EXPORT_SYMBOL(sync_blockdev); | |
287 | ||
288 | /* | |
289 | * Write out and wait upon all dirty data associated with this | |
290 | * device. Filesystem data as well as the underlying block | |
291 | * device. Takes the superblock lock. | |
292 | */ | |
293 | int fsync_bdev(struct block_device *bdev) | |
294 | { | |
295 | struct super_block *sb = get_super(bdev); | |
296 | if (sb) { | |
60b0680f | 297 | int res = sync_filesystem(sb); |
585d3bc0 NP |
298 | drop_super(sb); |
299 | return res; | |
300 | } | |
301 | return sync_blockdev(bdev); | |
302 | } | |
47e4491b | 303 | EXPORT_SYMBOL(fsync_bdev); |
585d3bc0 NP |
304 | |
305 | /** | |
306 | * freeze_bdev -- lock a filesystem and force it into a consistent state | |
307 | * @bdev: blockdevice to lock | |
308 | * | |
585d3bc0 NP |
309 | * If a superblock is found on this device, we take the s_umount semaphore |
310 | * on it to make sure nobody unmounts until the snapshot creation is done. | |
311 | * The reference counter (bd_fsfreeze_count) guarantees that only the last | |
312 | * unfreeze process can unfreeze the frozen filesystem actually when multiple | |
313 | * freeze requests arrive simultaneously. It counts up in freeze_bdev() and | |
314 | * count down in thaw_bdev(). When it becomes 0, thaw_bdev() will unfreeze | |
315 | * actually. | |
316 | */ | |
317 | struct super_block *freeze_bdev(struct block_device *bdev) | |
318 | { | |
319 | struct super_block *sb; | |
320 | int error = 0; | |
321 | ||
322 | mutex_lock(&bdev->bd_fsfreeze_mutex); | |
4504230a CH |
323 | if (++bdev->bd_fsfreeze_count > 1) { |
324 | /* | |
325 | * We don't even need to grab a reference - the first call | |
326 | * to freeze_bdev grab an active reference and only the last | |
327 | * thaw_bdev drops it. | |
328 | */ | |
585d3bc0 | 329 | sb = get_super(bdev); |
5bb53c0f AR |
330 | if (sb) |
331 | drop_super(sb); | |
4504230a CH |
332 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
333 | return sb; | |
334 | } | |
335 | ||
336 | sb = get_active_super(bdev); | |
337 | if (!sb) | |
338 | goto out; | |
48b6bca6 BM |
339 | if (sb->s_op->freeze_super) |
340 | error = sb->s_op->freeze_super(sb); | |
341 | else | |
342 | error = freeze_super(sb); | |
18e9e510 JB |
343 | if (error) { |
344 | deactivate_super(sb); | |
345 | bdev->bd_fsfreeze_count--; | |
585d3bc0 | 346 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
18e9e510 | 347 | return ERR_PTR(error); |
585d3bc0 | 348 | } |
18e9e510 | 349 | deactivate_super(sb); |
4504230a | 350 | out: |
585d3bc0 NP |
351 | sync_blockdev(bdev); |
352 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
4fadd7bb | 353 | return sb; /* thaw_bdev releases s->s_umount */ |
585d3bc0 NP |
354 | } |
355 | EXPORT_SYMBOL(freeze_bdev); | |
356 | ||
357 | /** | |
358 | * thaw_bdev -- unlock filesystem | |
359 | * @bdev: blockdevice to unlock | |
360 | * @sb: associated superblock | |
361 | * | |
362 | * Unlocks the filesystem and marks it writeable again after freeze_bdev(). | |
363 | */ | |
364 | int thaw_bdev(struct block_device *bdev, struct super_block *sb) | |
365 | { | |
4504230a | 366 | int error = -EINVAL; |
585d3bc0 NP |
367 | |
368 | mutex_lock(&bdev->bd_fsfreeze_mutex); | |
4504230a | 369 | if (!bdev->bd_fsfreeze_count) |
18e9e510 | 370 | goto out; |
4504230a CH |
371 | |
372 | error = 0; | |
373 | if (--bdev->bd_fsfreeze_count > 0) | |
18e9e510 | 374 | goto out; |
4504230a CH |
375 | |
376 | if (!sb) | |
18e9e510 | 377 | goto out; |
4504230a | 378 | |
48b6bca6 BM |
379 | if (sb->s_op->thaw_super) |
380 | error = sb->s_op->thaw_super(sb); | |
381 | else | |
382 | error = thaw_super(sb); | |
997198ba | 383 | if (error) |
18e9e510 | 384 | bdev->bd_fsfreeze_count++; |
18e9e510 | 385 | out: |
585d3bc0 | 386 | mutex_unlock(&bdev->bd_fsfreeze_mutex); |
997198ba | 387 | return error; |
585d3bc0 NP |
388 | } |
389 | EXPORT_SYMBOL(thaw_bdev); | |
390 | ||
1da177e4 LT |
391 | static int blkdev_writepage(struct page *page, struct writeback_control *wbc) |
392 | { | |
393 | return block_write_full_page(page, blkdev_get_block, wbc); | |
394 | } | |
395 | ||
396 | static int blkdev_readpage(struct file * file, struct page * page) | |
397 | { | |
398 | return block_read_full_page(page, blkdev_get_block); | |
399 | } | |
400 | ||
447f05bb AM |
401 | static int blkdev_readpages(struct file *file, struct address_space *mapping, |
402 | struct list_head *pages, unsigned nr_pages) | |
403 | { | |
404 | return mpage_readpages(mapping, pages, nr_pages, blkdev_get_block); | |
405 | } | |
406 | ||
6272b5a5 NP |
407 | static int blkdev_write_begin(struct file *file, struct address_space *mapping, |
408 | loff_t pos, unsigned len, unsigned flags, | |
409 | struct page **pagep, void **fsdata) | |
1da177e4 | 410 | { |
155130a4 CH |
411 | return block_write_begin(mapping, pos, len, flags, pagep, |
412 | blkdev_get_block); | |
1da177e4 LT |
413 | } |
414 | ||
6272b5a5 NP |
415 | static int blkdev_write_end(struct file *file, struct address_space *mapping, |
416 | loff_t pos, unsigned len, unsigned copied, | |
417 | struct page *page, void *fsdata) | |
1da177e4 | 418 | { |
6272b5a5 NP |
419 | int ret; |
420 | ret = block_write_end(file, mapping, pos, len, copied, page, fsdata); | |
421 | ||
422 | unlock_page(page); | |
09cbfeaf | 423 | put_page(page); |
6272b5a5 NP |
424 | |
425 | return ret; | |
1da177e4 LT |
426 | } |
427 | ||
428 | /* | |
429 | * private llseek: | |
496ad9aa | 430 | * for a block special file file_inode(file)->i_size is zero |
1da177e4 LT |
431 | * so we compute the size by hand (just as in block_read/write above) |
432 | */ | |
965c8e59 | 433 | static loff_t block_llseek(struct file *file, loff_t offset, int whence) |
1da177e4 | 434 | { |
4ebb16ca | 435 | struct inode *bd_inode = bdev_file_inode(file); |
1da177e4 LT |
436 | loff_t retval; |
437 | ||
5955102c | 438 | inode_lock(bd_inode); |
5d48f3a2 | 439 | retval = fixed_size_llseek(file, offset, whence, i_size_read(bd_inode)); |
5955102c | 440 | inode_unlock(bd_inode); |
1da177e4 LT |
441 | return retval; |
442 | } | |
443 | ||
02c24a82 | 444 | int blkdev_fsync(struct file *filp, loff_t start, loff_t end, int datasync) |
1da177e4 | 445 | { |
4ebb16ca | 446 | struct inode *bd_inode = bdev_file_inode(filp); |
b8af67e2 | 447 | struct block_device *bdev = I_BDEV(bd_inode); |
ab0a9735 | 448 | int error; |
da5aa861 RW |
449 | |
450 | error = filemap_write_and_wait_range(filp->f_mapping, start, end); | |
451 | if (error) | |
452 | return error; | |
ab0a9735 | 453 | |
b8af67e2 AB |
454 | /* |
455 | * There is no need to serialise calls to blkdev_issue_flush with | |
456 | * i_mutex and doing so causes performance issues with concurrent | |
457 | * O_SYNC writers to a block device. | |
458 | */ | |
dd3932ed | 459 | error = blkdev_issue_flush(bdev, GFP_KERNEL, NULL); |
ab0a9735 CH |
460 | if (error == -EOPNOTSUPP) |
461 | error = 0; | |
b8af67e2 | 462 | |
ab0a9735 | 463 | return error; |
1da177e4 | 464 | } |
b1dd3b28 | 465 | EXPORT_SYMBOL(blkdev_fsync); |
1da177e4 | 466 | |
47a191fd MW |
467 | /** |
468 | * bdev_read_page() - Start reading a page from a block device | |
469 | * @bdev: The device to read the page from | |
470 | * @sector: The offset on the device to read the page to (need not be aligned) | |
471 | * @page: The page to read | |
472 | * | |
473 | * On entry, the page should be locked. It will be unlocked when the page | |
474 | * has been read. If the block driver implements rw_page synchronously, | |
475 | * that will be true on exit from this function, but it need not be. | |
476 | * | |
477 | * Errors returned by this function are usually "soft", eg out of memory, or | |
478 | * queue full; callers should try a different route to read this page rather | |
479 | * than propagate an error back up the stack. | |
480 | * | |
481 | * Return: negative errno if an error occurs, 0 if submission was successful. | |
482 | */ | |
483 | int bdev_read_page(struct block_device *bdev, sector_t sector, | |
484 | struct page *page) | |
485 | { | |
486 | const struct block_device_operations *ops = bdev->bd_disk->fops; | |
2e6edc95 DW |
487 | int result = -EOPNOTSUPP; |
488 | ||
f68eb1e7 | 489 | if (!ops->rw_page || bdev_get_integrity(bdev)) |
2e6edc95 DW |
490 | return result; |
491 | ||
6f3b0e8b | 492 | result = blk_queue_enter(bdev->bd_queue, false); |
2e6edc95 DW |
493 | if (result) |
494 | return result; | |
c11f0c0b | 495 | result = ops->rw_page(bdev, sector + get_start_sect(bdev), page, false); |
2e6edc95 DW |
496 | blk_queue_exit(bdev->bd_queue); |
497 | return result; | |
47a191fd MW |
498 | } |
499 | EXPORT_SYMBOL_GPL(bdev_read_page); | |
500 | ||
501 | /** | |
502 | * bdev_write_page() - Start writing a page to a block device | |
503 | * @bdev: The device to write the page to | |
504 | * @sector: The offset on the device to write the page to (need not be aligned) | |
505 | * @page: The page to write | |
506 | * @wbc: The writeback_control for the write | |
507 | * | |
508 | * On entry, the page should be locked and not currently under writeback. | |
509 | * On exit, if the write started successfully, the page will be unlocked and | |
510 | * under writeback. If the write failed already (eg the driver failed to | |
511 | * queue the page to the device), the page will still be locked. If the | |
512 | * caller is a ->writepage implementation, it will need to unlock the page. | |
513 | * | |
514 | * Errors returned by this function are usually "soft", eg out of memory, or | |
515 | * queue full; callers should try a different route to write this page rather | |
516 | * than propagate an error back up the stack. | |
517 | * | |
518 | * Return: negative errno if an error occurs, 0 if submission was successful. | |
519 | */ | |
520 | int bdev_write_page(struct block_device *bdev, sector_t sector, | |
521 | struct page *page, struct writeback_control *wbc) | |
522 | { | |
523 | int result; | |
47a191fd | 524 | const struct block_device_operations *ops = bdev->bd_disk->fops; |
2e6edc95 | 525 | |
f68eb1e7 | 526 | if (!ops->rw_page || bdev_get_integrity(bdev)) |
47a191fd | 527 | return -EOPNOTSUPP; |
6f3b0e8b | 528 | result = blk_queue_enter(bdev->bd_queue, false); |
2e6edc95 DW |
529 | if (result) |
530 | return result; | |
531 | ||
47a191fd | 532 | set_page_writeback(page); |
c11f0c0b | 533 | result = ops->rw_page(bdev, sector + get_start_sect(bdev), page, true); |
47a191fd MW |
534 | if (result) |
535 | end_page_writeback(page); | |
536 | else | |
537 | unlock_page(page); | |
2e6edc95 | 538 | blk_queue_exit(bdev->bd_queue); |
47a191fd MW |
539 | return result; |
540 | } | |
541 | EXPORT_SYMBOL_GPL(bdev_write_page); | |
542 | ||
dd22f551 MW |
543 | /** |
544 | * bdev_direct_access() - Get the address for directly-accessibly memory | |
545 | * @bdev: The device containing the memory | |
b2e0d162 | 546 | * @dax: control and output parameters for ->direct_access |
dd22f551 MW |
547 | * |
548 | * If a block device is made up of directly addressable memory, this function | |
549 | * will tell the caller the PFN and the address of the memory. The address | |
550 | * may be directly dereferenced within the kernel without the need to call | |
551 | * ioremap(), kmap() or similar. The PFN is suitable for inserting into | |
552 | * page tables. | |
553 | * | |
554 | * Return: negative errno if an error occurs, otherwise the number of bytes | |
555 | * accessible at this address. | |
556 | */ | |
b2e0d162 | 557 | long bdev_direct_access(struct block_device *bdev, struct blk_dax_ctl *dax) |
dd22f551 | 558 | { |
b2e0d162 DW |
559 | sector_t sector = dax->sector; |
560 | long avail, size = dax->size; | |
dd22f551 MW |
561 | const struct block_device_operations *ops = bdev->bd_disk->fops; |
562 | ||
43c3dd08 MW |
563 | /* |
564 | * The device driver is allowed to sleep, in order to make the | |
565 | * memory directly accessible. | |
566 | */ | |
567 | might_sleep(); | |
568 | ||
dd22f551 MW |
569 | if (size < 0) |
570 | return size; | |
163d4baa | 571 | if (!blk_queue_dax(bdev_get_queue(bdev)) || !ops->direct_access) |
dd22f551 MW |
572 | return -EOPNOTSUPP; |
573 | if ((sector + DIV_ROUND_UP(size, 512)) > | |
574 | part_nr_sects_read(bdev->bd_part)) | |
575 | return -ERANGE; | |
576 | sector += get_start_sect(bdev); | |
577 | if (sector % (PAGE_SIZE / 512)) | |
578 | return -EINVAL; | |
0a70bd43 | 579 | avail = ops->direct_access(bdev, sector, &dax->addr, &dax->pfn, size); |
dd22f551 MW |
580 | if (!avail) |
581 | return -ERANGE; | |
fe683ada DW |
582 | if (avail > 0 && avail & ~PAGE_MASK) |
583 | return -ENXIO; | |
dd22f551 MW |
584 | return min(avail, size); |
585 | } | |
586 | EXPORT_SYMBOL_GPL(bdev_direct_access); | |
587 | ||
2d96afc8 TK |
588 | /** |
589 | * bdev_dax_supported() - Check if the device supports dax for filesystem | |
590 | * @sb: The superblock of the device | |
591 | * @blocksize: The block size of the device | |
592 | * | |
593 | * This is a library function for filesystems to check if the block device | |
594 | * can be mounted with dax option. | |
595 | * | |
596 | * Return: negative errno if unsupported, 0 if supported. | |
597 | */ | |
598 | int bdev_dax_supported(struct super_block *sb, int blocksize) | |
599 | { | |
600 | struct blk_dax_ctl dax = { | |
601 | .sector = 0, | |
602 | .size = PAGE_SIZE, | |
603 | }; | |
604 | int err; | |
605 | ||
606 | if (blocksize != PAGE_SIZE) { | |
607 | vfs_msg(sb, KERN_ERR, "error: unsupported blocksize for dax"); | |
608 | return -EINVAL; | |
609 | } | |
610 | ||
611 | err = bdev_direct_access(sb->s_bdev, &dax); | |
612 | if (err < 0) { | |
613 | switch (err) { | |
614 | case -EOPNOTSUPP: | |
615 | vfs_msg(sb, KERN_ERR, | |
616 | "error: device does not support dax"); | |
617 | break; | |
618 | case -EINVAL: | |
619 | vfs_msg(sb, KERN_ERR, | |
620 | "error: unaligned partition for dax"); | |
621 | break; | |
622 | default: | |
623 | vfs_msg(sb, KERN_ERR, | |
624 | "error: dax access failed (%d)", err); | |
625 | } | |
626 | return err; | |
627 | } | |
628 | ||
629 | return 0; | |
630 | } | |
631 | EXPORT_SYMBOL_GPL(bdev_dax_supported); | |
632 | ||
a8078b1f TK |
633 | /** |
634 | * bdev_dax_capable() - Return if the raw device is capable for dax | |
635 | * @bdev: The device for raw block device access | |
636 | */ | |
637 | bool bdev_dax_capable(struct block_device *bdev) | |
638 | { | |
a8078b1f TK |
639 | struct blk_dax_ctl dax = { |
640 | .size = PAGE_SIZE, | |
641 | }; | |
642 | ||
643 | if (!IS_ENABLED(CONFIG_FS_DAX)) | |
644 | return false; | |
645 | ||
646 | dax.sector = 0; | |
647 | if (bdev_direct_access(bdev, &dax) < 0) | |
648 | return false; | |
649 | ||
650 | dax.sector = bdev->bd_part->nr_sects - (PAGE_SIZE / 512); | |
651 | if (bdev_direct_access(bdev, &dax) < 0) | |
652 | return false; | |
653 | ||
a8078b1f TK |
654 | return true; |
655 | } | |
656 | ||
1da177e4 LT |
657 | /* |
658 | * pseudo-fs | |
659 | */ | |
660 | ||
661 | static __cacheline_aligned_in_smp DEFINE_SPINLOCK(bdev_lock); | |
e18b890b | 662 | static struct kmem_cache * bdev_cachep __read_mostly; |
1da177e4 LT |
663 | |
664 | static struct inode *bdev_alloc_inode(struct super_block *sb) | |
665 | { | |
e94b1766 | 666 | struct bdev_inode *ei = kmem_cache_alloc(bdev_cachep, GFP_KERNEL); |
1da177e4 LT |
667 | if (!ei) |
668 | return NULL; | |
669 | return &ei->vfs_inode; | |
670 | } | |
671 | ||
fa0d7e3d | 672 | static void bdev_i_callback(struct rcu_head *head) |
1da177e4 | 673 | { |
fa0d7e3d | 674 | struct inode *inode = container_of(head, struct inode, i_rcu); |
1da177e4 LT |
675 | struct bdev_inode *bdi = BDEV_I(inode); |
676 | ||
1da177e4 LT |
677 | kmem_cache_free(bdev_cachep, bdi); |
678 | } | |
679 | ||
fa0d7e3d NP |
680 | static void bdev_destroy_inode(struct inode *inode) |
681 | { | |
682 | call_rcu(&inode->i_rcu, bdev_i_callback); | |
683 | } | |
684 | ||
51cc5068 | 685 | static void init_once(void *foo) |
1da177e4 LT |
686 | { |
687 | struct bdev_inode *ei = (struct bdev_inode *) foo; | |
688 | struct block_device *bdev = &ei->bdev; | |
689 | ||
a35afb83 CL |
690 | memset(bdev, 0, sizeof(*bdev)); |
691 | mutex_init(&bdev->bd_mutex); | |
a35afb83 | 692 | INIT_LIST_HEAD(&bdev->bd_list); |
49731baa TH |
693 | #ifdef CONFIG_SYSFS |
694 | INIT_LIST_HEAD(&bdev->bd_holder_disks); | |
695 | #endif | |
a35afb83 | 696 | inode_init_once(&ei->vfs_inode); |
fcccf502 TS |
697 | /* Initialize mutex for freeze. */ |
698 | mutex_init(&bdev->bd_fsfreeze_mutex); | |
1da177e4 LT |
699 | } |
700 | ||
b57922d9 | 701 | static void bdev_evict_inode(struct inode *inode) |
1da177e4 LT |
702 | { |
703 | struct block_device *bdev = &BDEV_I(inode)->bdev; | |
91b0abe3 | 704 | truncate_inode_pages_final(&inode->i_data); |
b57922d9 | 705 | invalidate_inode_buffers(inode); /* is it needed here? */ |
dbd5768f | 706 | clear_inode(inode); |
1da177e4 | 707 | spin_lock(&bdev_lock); |
1da177e4 LT |
708 | list_del_init(&bdev->bd_list); |
709 | spin_unlock(&bdev_lock); | |
710 | } | |
711 | ||
ee9b6d61 | 712 | static const struct super_operations bdev_sops = { |
1da177e4 LT |
713 | .statfs = simple_statfs, |
714 | .alloc_inode = bdev_alloc_inode, | |
715 | .destroy_inode = bdev_destroy_inode, | |
716 | .drop_inode = generic_delete_inode, | |
b57922d9 | 717 | .evict_inode = bdev_evict_inode, |
1da177e4 LT |
718 | }; |
719 | ||
51139ada AV |
720 | static struct dentry *bd_mount(struct file_system_type *fs_type, |
721 | int flags, const char *dev_name, void *data) | |
1da177e4 | 722 | { |
3684aa70 SL |
723 | struct dentry *dent; |
724 | dent = mount_pseudo(fs_type, "bdev:", &bdev_sops, NULL, BDEVFS_MAGIC); | |
e9e5e3fa | 725 | if (!IS_ERR(dent)) |
3684aa70 SL |
726 | dent->d_sb->s_iflags |= SB_I_CGROUPWB; |
727 | return dent; | |
1da177e4 LT |
728 | } |
729 | ||
730 | static struct file_system_type bd_type = { | |
731 | .name = "bdev", | |
51139ada | 732 | .mount = bd_mount, |
1da177e4 LT |
733 | .kill_sb = kill_anon_super, |
734 | }; | |
735 | ||
a212b105 TH |
736 | struct super_block *blockdev_superblock __read_mostly; |
737 | EXPORT_SYMBOL_GPL(blockdev_superblock); | |
1da177e4 LT |
738 | |
739 | void __init bdev_cache_init(void) | |
740 | { | |
741 | int err; | |
ace8577a | 742 | static struct vfsmount *bd_mnt; |
c2acf7b9 | 743 | |
1da177e4 | 744 | bdev_cachep = kmem_cache_create("bdev_cache", sizeof(struct bdev_inode), |
fffb60f9 | 745 | 0, (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT| |
5d097056 | 746 | SLAB_MEM_SPREAD|SLAB_ACCOUNT|SLAB_PANIC), |
20c2df83 | 747 | init_once); |
1da177e4 LT |
748 | err = register_filesystem(&bd_type); |
749 | if (err) | |
750 | panic("Cannot register bdev pseudo-fs"); | |
751 | bd_mnt = kern_mount(&bd_type); | |
1da177e4 LT |
752 | if (IS_ERR(bd_mnt)) |
753 | panic("Cannot create bdev pseudo-fs"); | |
ace8577a | 754 | blockdev_superblock = bd_mnt->mnt_sb; /* For writeback */ |
1da177e4 LT |
755 | } |
756 | ||
757 | /* | |
758 | * Most likely _very_ bad one - but then it's hardly critical for small | |
759 | * /dev and can be fixed when somebody will need really large one. | |
760 | * Keep in mind that it will be fed through icache hash function too. | |
761 | */ | |
762 | static inline unsigned long hash(dev_t dev) | |
763 | { | |
764 | return MAJOR(dev)+MINOR(dev); | |
765 | } | |
766 | ||
767 | static int bdev_test(struct inode *inode, void *data) | |
768 | { | |
769 | return BDEV_I(inode)->bdev.bd_dev == *(dev_t *)data; | |
770 | } | |
771 | ||
772 | static int bdev_set(struct inode *inode, void *data) | |
773 | { | |
774 | BDEV_I(inode)->bdev.bd_dev = *(dev_t *)data; | |
775 | return 0; | |
776 | } | |
777 | ||
778 | static LIST_HEAD(all_bdevs); | |
779 | ||
780 | struct block_device *bdget(dev_t dev) | |
781 | { | |
782 | struct block_device *bdev; | |
783 | struct inode *inode; | |
784 | ||
c2acf7b9 | 785 | inode = iget5_locked(blockdev_superblock, hash(dev), |
1da177e4 LT |
786 | bdev_test, bdev_set, &dev); |
787 | ||
788 | if (!inode) | |
789 | return NULL; | |
790 | ||
791 | bdev = &BDEV_I(inode)->bdev; | |
792 | ||
793 | if (inode->i_state & I_NEW) { | |
794 | bdev->bd_contains = NULL; | |
782b94cd | 795 | bdev->bd_super = NULL; |
1da177e4 LT |
796 | bdev->bd_inode = inode; |
797 | bdev->bd_block_size = (1 << inode->i_blkbits); | |
798 | bdev->bd_part_count = 0; | |
799 | bdev->bd_invalidated = 0; | |
800 | inode->i_mode = S_IFBLK; | |
801 | inode->i_rdev = dev; | |
802 | inode->i_bdev = bdev; | |
803 | inode->i_data.a_ops = &def_blk_aops; | |
804 | mapping_set_gfp_mask(&inode->i_data, GFP_USER); | |
1da177e4 LT |
805 | spin_lock(&bdev_lock); |
806 | list_add(&bdev->bd_list, &all_bdevs); | |
807 | spin_unlock(&bdev_lock); | |
808 | unlock_new_inode(inode); | |
809 | } | |
810 | return bdev; | |
811 | } | |
812 | ||
813 | EXPORT_SYMBOL(bdget); | |
814 | ||
dddac6a7 AJ |
815 | /** |
816 | * bdgrab -- Grab a reference to an already referenced block device | |
817 | * @bdev: Block device to grab a reference to. | |
818 | */ | |
819 | struct block_device *bdgrab(struct block_device *bdev) | |
820 | { | |
7de9c6ee | 821 | ihold(bdev->bd_inode); |
dddac6a7 AJ |
822 | return bdev; |
823 | } | |
c1681bf8 | 824 | EXPORT_SYMBOL(bdgrab); |
dddac6a7 | 825 | |
1da177e4 LT |
826 | long nr_blockdev_pages(void) |
827 | { | |
203a2935 | 828 | struct block_device *bdev; |
1da177e4 LT |
829 | long ret = 0; |
830 | spin_lock(&bdev_lock); | |
203a2935 | 831 | list_for_each_entry(bdev, &all_bdevs, bd_list) { |
1da177e4 LT |
832 | ret += bdev->bd_inode->i_mapping->nrpages; |
833 | } | |
834 | spin_unlock(&bdev_lock); | |
835 | return ret; | |
836 | } | |
837 | ||
838 | void bdput(struct block_device *bdev) | |
839 | { | |
840 | iput(bdev->bd_inode); | |
841 | } | |
842 | ||
843 | EXPORT_SYMBOL(bdput); | |
844 | ||
845 | static struct block_device *bd_acquire(struct inode *inode) | |
846 | { | |
847 | struct block_device *bdev; | |
09d967c6 | 848 | |
1da177e4 LT |
849 | spin_lock(&bdev_lock); |
850 | bdev = inode->i_bdev; | |
09d967c6 | 851 | if (bdev) { |
ed8a9d2c | 852 | bdgrab(bdev); |
1da177e4 LT |
853 | spin_unlock(&bdev_lock); |
854 | return bdev; | |
855 | } | |
856 | spin_unlock(&bdev_lock); | |
09d967c6 | 857 | |
1da177e4 LT |
858 | bdev = bdget(inode->i_rdev); |
859 | if (bdev) { | |
860 | spin_lock(&bdev_lock); | |
09d967c6 OH |
861 | if (!inode->i_bdev) { |
862 | /* | |
7de9c6ee | 863 | * We take an additional reference to bd_inode, |
09d967c6 OH |
864 | * and it's released in clear_inode() of inode. |
865 | * So, we can access it via ->i_mapping always | |
866 | * without igrab(). | |
867 | */ | |
ed8a9d2c | 868 | bdgrab(bdev); |
09d967c6 OH |
869 | inode->i_bdev = bdev; |
870 | inode->i_mapping = bdev->bd_inode->i_mapping; | |
09d967c6 | 871 | } |
1da177e4 LT |
872 | spin_unlock(&bdev_lock); |
873 | } | |
874 | return bdev; | |
875 | } | |
876 | ||
877 | /* Call when you free inode */ | |
878 | ||
879 | void bd_forget(struct inode *inode) | |
880 | { | |
09d967c6 OH |
881 | struct block_device *bdev = NULL; |
882 | ||
1da177e4 | 883 | spin_lock(&bdev_lock); |
b4ea2eaa YH |
884 | if (!sb_is_blkdev_sb(inode->i_sb)) |
885 | bdev = inode->i_bdev; | |
a4a4f943 AV |
886 | inode->i_bdev = NULL; |
887 | inode->i_mapping = &inode->i_data; | |
1da177e4 | 888 | spin_unlock(&bdev_lock); |
09d967c6 OH |
889 | |
890 | if (bdev) | |
ed8a9d2c | 891 | bdput(bdev); |
1da177e4 LT |
892 | } |
893 | ||
1a3cbbc5 TH |
894 | /** |
895 | * bd_may_claim - test whether a block device can be claimed | |
896 | * @bdev: block device of interest | |
897 | * @whole: whole block device containing @bdev, may equal @bdev | |
898 | * @holder: holder trying to claim @bdev | |
899 | * | |
25985edc | 900 | * Test whether @bdev can be claimed by @holder. |
1a3cbbc5 TH |
901 | * |
902 | * CONTEXT: | |
903 | * spin_lock(&bdev_lock). | |
904 | * | |
905 | * RETURNS: | |
906 | * %true if @bdev can be claimed, %false otherwise. | |
907 | */ | |
908 | static bool bd_may_claim(struct block_device *bdev, struct block_device *whole, | |
909 | void *holder) | |
1da177e4 | 910 | { |
1da177e4 | 911 | if (bdev->bd_holder == holder) |
1a3cbbc5 | 912 | return true; /* already a holder */ |
1da177e4 | 913 | else if (bdev->bd_holder != NULL) |
1a3cbbc5 | 914 | return false; /* held by someone else */ |
1da177e4 | 915 | else if (bdev->bd_contains == bdev) |
1a3cbbc5 | 916 | return true; /* is a whole device which isn't held */ |
1da177e4 | 917 | |
e525fd89 | 918 | else if (whole->bd_holder == bd_may_claim) |
1a3cbbc5 TH |
919 | return true; /* is a partition of a device that is being partitioned */ |
920 | else if (whole->bd_holder != NULL) | |
921 | return false; /* is a partition of a held device */ | |
1da177e4 | 922 | else |
1a3cbbc5 TH |
923 | return true; /* is a partition of an un-held device */ |
924 | } | |
925 | ||
6b4517a7 TH |
926 | /** |
927 | * bd_prepare_to_claim - prepare to claim a block device | |
928 | * @bdev: block device of interest | |
929 | * @whole: the whole device containing @bdev, may equal @bdev | |
930 | * @holder: holder trying to claim @bdev | |
931 | * | |
932 | * Prepare to claim @bdev. This function fails if @bdev is already | |
933 | * claimed by another holder and waits if another claiming is in | |
934 | * progress. This function doesn't actually claim. On successful | |
935 | * return, the caller has ownership of bd_claiming and bd_holder[s]. | |
936 | * | |
937 | * CONTEXT: | |
938 | * spin_lock(&bdev_lock). Might release bdev_lock, sleep and regrab | |
939 | * it multiple times. | |
940 | * | |
941 | * RETURNS: | |
942 | * 0 if @bdev can be claimed, -EBUSY otherwise. | |
943 | */ | |
944 | static int bd_prepare_to_claim(struct block_device *bdev, | |
945 | struct block_device *whole, void *holder) | |
946 | { | |
947 | retry: | |
948 | /* if someone else claimed, fail */ | |
949 | if (!bd_may_claim(bdev, whole, holder)) | |
950 | return -EBUSY; | |
951 | ||
e75aa858 TH |
952 | /* if claiming is already in progress, wait for it to finish */ |
953 | if (whole->bd_claiming) { | |
6b4517a7 TH |
954 | wait_queue_head_t *wq = bit_waitqueue(&whole->bd_claiming, 0); |
955 | DEFINE_WAIT(wait); | |
956 | ||
957 | prepare_to_wait(wq, &wait, TASK_UNINTERRUPTIBLE); | |
958 | spin_unlock(&bdev_lock); | |
959 | schedule(); | |
960 | finish_wait(wq, &wait); | |
961 | spin_lock(&bdev_lock); | |
962 | goto retry; | |
963 | } | |
964 | ||
965 | /* yay, all mine */ | |
966 | return 0; | |
967 | } | |
968 | ||
969 | /** | |
970 | * bd_start_claiming - start claiming a block device | |
971 | * @bdev: block device of interest | |
972 | * @holder: holder trying to claim @bdev | |
973 | * | |
974 | * @bdev is about to be opened exclusively. Check @bdev can be opened | |
975 | * exclusively and mark that an exclusive open is in progress. Each | |
976 | * successful call to this function must be matched with a call to | |
b0018361 NP |
977 | * either bd_finish_claiming() or bd_abort_claiming() (which do not |
978 | * fail). | |
979 | * | |
980 | * This function is used to gain exclusive access to the block device | |
981 | * without actually causing other exclusive open attempts to fail. It | |
982 | * should be used when the open sequence itself requires exclusive | |
983 | * access but may subsequently fail. | |
6b4517a7 TH |
984 | * |
985 | * CONTEXT: | |
986 | * Might sleep. | |
987 | * | |
988 | * RETURNS: | |
989 | * Pointer to the block device containing @bdev on success, ERR_PTR() | |
990 | * value on failure. | |
991 | */ | |
992 | static struct block_device *bd_start_claiming(struct block_device *bdev, | |
993 | void *holder) | |
994 | { | |
995 | struct gendisk *disk; | |
996 | struct block_device *whole; | |
997 | int partno, err; | |
998 | ||
999 | might_sleep(); | |
1000 | ||
1001 | /* | |
1002 | * @bdev might not have been initialized properly yet, look up | |
1003 | * and grab the outer block device the hard way. | |
1004 | */ | |
1005 | disk = get_gendisk(bdev->bd_dev, &partno); | |
1006 | if (!disk) | |
1007 | return ERR_PTR(-ENXIO); | |
1008 | ||
d4c208b8 TH |
1009 | /* |
1010 | * Normally, @bdev should equal what's returned from bdget_disk() | |
1011 | * if partno is 0; however, some drivers (floppy) use multiple | |
1012 | * bdev's for the same physical device and @bdev may be one of the | |
1013 | * aliases. Keep @bdev if partno is 0. This means claimer | |
1014 | * tracking is broken for those devices but it has always been that | |
1015 | * way. | |
1016 | */ | |
1017 | if (partno) | |
1018 | whole = bdget_disk(disk, 0); | |
1019 | else | |
1020 | whole = bdgrab(bdev); | |
1021 | ||
cf342570 | 1022 | module_put(disk->fops->owner); |
6b4517a7 TH |
1023 | put_disk(disk); |
1024 | if (!whole) | |
1025 | return ERR_PTR(-ENOMEM); | |
1026 | ||
1027 | /* prepare to claim, if successful, mark claiming in progress */ | |
1028 | spin_lock(&bdev_lock); | |
1029 | ||
1030 | err = bd_prepare_to_claim(bdev, whole, holder); | |
1031 | if (err == 0) { | |
1032 | whole->bd_claiming = holder; | |
1033 | spin_unlock(&bdev_lock); | |
1034 | return whole; | |
1035 | } else { | |
1036 | spin_unlock(&bdev_lock); | |
1037 | bdput(whole); | |
1038 | return ERR_PTR(err); | |
1039 | } | |
1040 | } | |
1041 | ||
641dc636 | 1042 | #ifdef CONFIG_SYSFS |
49731baa TH |
1043 | struct bd_holder_disk { |
1044 | struct list_head list; | |
1045 | struct gendisk *disk; | |
1046 | int refcnt; | |
1047 | }; | |
1048 | ||
1049 | static struct bd_holder_disk *bd_find_holder_disk(struct block_device *bdev, | |
1050 | struct gendisk *disk) | |
1051 | { | |
1052 | struct bd_holder_disk *holder; | |
1053 | ||
1054 | list_for_each_entry(holder, &bdev->bd_holder_disks, list) | |
1055 | if (holder->disk == disk) | |
1056 | return holder; | |
1057 | return NULL; | |
1058 | } | |
1059 | ||
4d7dd8fd | 1060 | static int add_symlink(struct kobject *from, struct kobject *to) |
641dc636 | 1061 | { |
4d7dd8fd | 1062 | return sysfs_create_link(from, to, kobject_name(to)); |
641dc636 JN |
1063 | } |
1064 | ||
1065 | static void del_symlink(struct kobject *from, struct kobject *to) | |
1066 | { | |
641dc636 JN |
1067 | sysfs_remove_link(from, kobject_name(to)); |
1068 | } | |
1069 | ||
df6c0cd9 | 1070 | /** |
e09b457b TH |
1071 | * bd_link_disk_holder - create symlinks between holding disk and slave bdev |
1072 | * @bdev: the claimed slave bdev | |
1073 | * @disk: the holding disk | |
df6c0cd9 | 1074 | * |
49731baa TH |
1075 | * DON'T USE THIS UNLESS YOU'RE ALREADY USING IT. |
1076 | * | |
e09b457b | 1077 | * This functions creates the following sysfs symlinks. |
641dc636 | 1078 | * |
e09b457b TH |
1079 | * - from "slaves" directory of the holder @disk to the claimed @bdev |
1080 | * - from "holders" directory of the @bdev to the holder @disk | |
641dc636 | 1081 | * |
e09b457b TH |
1082 | * For example, if /dev/dm-0 maps to /dev/sda and disk for dm-0 is |
1083 | * passed to bd_link_disk_holder(), then: | |
641dc636 | 1084 | * |
e09b457b TH |
1085 | * /sys/block/dm-0/slaves/sda --> /sys/block/sda |
1086 | * /sys/block/sda/holders/dm-0 --> /sys/block/dm-0 | |
641dc636 | 1087 | * |
e09b457b TH |
1088 | * The caller must have claimed @bdev before calling this function and |
1089 | * ensure that both @bdev and @disk are valid during the creation and | |
1090 | * lifetime of these symlinks. | |
641dc636 | 1091 | * |
e09b457b TH |
1092 | * CONTEXT: |
1093 | * Might sleep. | |
641dc636 | 1094 | * |
e09b457b TH |
1095 | * RETURNS: |
1096 | * 0 on success, -errno on failure. | |
641dc636 | 1097 | */ |
e09b457b | 1098 | int bd_link_disk_holder(struct block_device *bdev, struct gendisk *disk) |
641dc636 | 1099 | { |
49731baa | 1100 | struct bd_holder_disk *holder; |
e09b457b | 1101 | int ret = 0; |
641dc636 | 1102 | |
2e7b651d | 1103 | mutex_lock(&bdev->bd_mutex); |
df6c0cd9 | 1104 | |
49731baa | 1105 | WARN_ON_ONCE(!bdev->bd_holder); |
4e91672c | 1106 | |
e09b457b TH |
1107 | /* FIXME: remove the following once add_disk() handles errors */ |
1108 | if (WARN_ON(!disk->slave_dir || !bdev->bd_part->holder_dir)) | |
1109 | goto out_unlock; | |
4e91672c | 1110 | |
49731baa TH |
1111 | holder = bd_find_holder_disk(bdev, disk); |
1112 | if (holder) { | |
1113 | holder->refcnt++; | |
e09b457b | 1114 | goto out_unlock; |
49731baa | 1115 | } |
641dc636 | 1116 | |
49731baa TH |
1117 | holder = kzalloc(sizeof(*holder), GFP_KERNEL); |
1118 | if (!holder) { | |
1119 | ret = -ENOMEM; | |
e09b457b TH |
1120 | goto out_unlock; |
1121 | } | |
641dc636 | 1122 | |
49731baa TH |
1123 | INIT_LIST_HEAD(&holder->list); |
1124 | holder->disk = disk; | |
1125 | holder->refcnt = 1; | |
1126 | ||
1127 | ret = add_symlink(disk->slave_dir, &part_to_dev(bdev->bd_part)->kobj); | |
1128 | if (ret) | |
1129 | goto out_free; | |
1130 | ||
1131 | ret = add_symlink(bdev->bd_part->holder_dir, &disk_to_dev(disk)->kobj); | |
1132 | if (ret) | |
1133 | goto out_del; | |
e7407d16 TH |
1134 | /* |
1135 | * bdev could be deleted beneath us which would implicitly destroy | |
1136 | * the holder directory. Hold on to it. | |
1137 | */ | |
1138 | kobject_get(bdev->bd_part->holder_dir); | |
49731baa TH |
1139 | |
1140 | list_add(&holder->list, &bdev->bd_holder_disks); | |
1141 | goto out_unlock; | |
1142 | ||
1143 | out_del: | |
1144 | del_symlink(disk->slave_dir, &part_to_dev(bdev->bd_part)->kobj); | |
1145 | out_free: | |
1146 | kfree(holder); | |
e09b457b | 1147 | out_unlock: |
b4cf1b72 | 1148 | mutex_unlock(&bdev->bd_mutex); |
e09b457b | 1149 | return ret; |
641dc636 | 1150 | } |
e09b457b | 1151 | EXPORT_SYMBOL_GPL(bd_link_disk_holder); |
641dc636 | 1152 | |
49731baa TH |
1153 | /** |
1154 | * bd_unlink_disk_holder - destroy symlinks created by bd_link_disk_holder() | |
1155 | * @bdev: the calimed slave bdev | |
1156 | * @disk: the holding disk | |
1157 | * | |
1158 | * DON'T USE THIS UNLESS YOU'RE ALREADY USING IT. | |
1159 | * | |
1160 | * CONTEXT: | |
1161 | * Might sleep. | |
1162 | */ | |
1163 | void bd_unlink_disk_holder(struct block_device *bdev, struct gendisk *disk) | |
641dc636 | 1164 | { |
49731baa | 1165 | struct bd_holder_disk *holder; |
641dc636 | 1166 | |
49731baa | 1167 | mutex_lock(&bdev->bd_mutex); |
641dc636 | 1168 | |
49731baa TH |
1169 | holder = bd_find_holder_disk(bdev, disk); |
1170 | ||
1171 | if (!WARN_ON_ONCE(holder == NULL) && !--holder->refcnt) { | |
1172 | del_symlink(disk->slave_dir, &part_to_dev(bdev->bd_part)->kobj); | |
1173 | del_symlink(bdev->bd_part->holder_dir, | |
1174 | &disk_to_dev(disk)->kobj); | |
e7407d16 | 1175 | kobject_put(bdev->bd_part->holder_dir); |
49731baa TH |
1176 | list_del_init(&holder->list); |
1177 | kfree(holder); | |
1178 | } | |
1179 | ||
1180 | mutex_unlock(&bdev->bd_mutex); | |
1da177e4 | 1181 | } |
49731baa | 1182 | EXPORT_SYMBOL_GPL(bd_unlink_disk_holder); |
641dc636 | 1183 | #endif |
1da177e4 | 1184 | |
56ade44b AP |
1185 | /** |
1186 | * flush_disk - invalidates all buffer-cache entries on a disk | |
1187 | * | |
1188 | * @bdev: struct block device to be flushed | |
e6eb5ce1 | 1189 | * @kill_dirty: flag to guide handling of dirty inodes |
56ade44b AP |
1190 | * |
1191 | * Invalidates all buffer-cache entries on a disk. It should be called | |
1192 | * when a disk has been changed -- either by a media change or online | |
1193 | * resize. | |
1194 | */ | |
93b270f7 | 1195 | static void flush_disk(struct block_device *bdev, bool kill_dirty) |
56ade44b | 1196 | { |
93b270f7 | 1197 | if (__invalidate_device(bdev, kill_dirty)) { |
56ade44b | 1198 | printk(KERN_WARNING "VFS: busy inodes on changed media or " |
424081f3 DM |
1199 | "resized disk %s\n", |
1200 | bdev->bd_disk ? bdev->bd_disk->disk_name : ""); | |
56ade44b AP |
1201 | } |
1202 | ||
1203 | if (!bdev->bd_disk) | |
1204 | return; | |
d27769ec | 1205 | if (disk_part_scan_enabled(bdev->bd_disk)) |
56ade44b AP |
1206 | bdev->bd_invalidated = 1; |
1207 | } | |
1208 | ||
c3279d14 | 1209 | /** |
57d1b536 | 1210 | * check_disk_size_change - checks for disk size change and adjusts bdev size. |
c3279d14 AP |
1211 | * @disk: struct gendisk to check |
1212 | * @bdev: struct bdev to adjust. | |
1213 | * | |
1214 | * This routine checks to see if the bdev size does not match the disk size | |
1215 | * and adjusts it if it differs. | |
1216 | */ | |
1217 | void check_disk_size_change(struct gendisk *disk, struct block_device *bdev) | |
1218 | { | |
1219 | loff_t disk_size, bdev_size; | |
1220 | ||
1221 | disk_size = (loff_t)get_capacity(disk) << 9; | |
1222 | bdev_size = i_size_read(bdev->bd_inode); | |
1223 | if (disk_size != bdev_size) { | |
c3279d14 AP |
1224 | printk(KERN_INFO |
1225 | "%s: detected capacity change from %lld to %lld\n", | |
424081f3 | 1226 | disk->disk_name, bdev_size, disk_size); |
c3279d14 | 1227 | i_size_write(bdev->bd_inode, disk_size); |
93b270f7 | 1228 | flush_disk(bdev, false); |
c3279d14 AP |
1229 | } |
1230 | } | |
1231 | EXPORT_SYMBOL(check_disk_size_change); | |
1232 | ||
0c002c2f | 1233 | /** |
57d1b536 | 1234 | * revalidate_disk - wrapper for lower-level driver's revalidate_disk call-back |
0c002c2f AP |
1235 | * @disk: struct gendisk to be revalidated |
1236 | * | |
1237 | * This routine is a wrapper for lower-level driver's revalidate_disk | |
1238 | * call-backs. It is used to do common pre and post operations needed | |
1239 | * for all revalidate_disk operations. | |
1240 | */ | |
1241 | int revalidate_disk(struct gendisk *disk) | |
1242 | { | |
c3279d14 | 1243 | struct block_device *bdev; |
0c002c2f AP |
1244 | int ret = 0; |
1245 | ||
1246 | if (disk->fops->revalidate_disk) | |
1247 | ret = disk->fops->revalidate_disk(disk); | |
25520d55 | 1248 | blk_integrity_revalidate(disk); |
c3279d14 AP |
1249 | bdev = bdget_disk(disk, 0); |
1250 | if (!bdev) | |
1251 | return ret; | |
1252 | ||
1253 | mutex_lock(&bdev->bd_mutex); | |
1254 | check_disk_size_change(disk, bdev); | |
7630b661 | 1255 | bdev->bd_invalidated = 0; |
c3279d14 AP |
1256 | mutex_unlock(&bdev->bd_mutex); |
1257 | bdput(bdev); | |
0c002c2f AP |
1258 | return ret; |
1259 | } | |
1260 | EXPORT_SYMBOL(revalidate_disk); | |
1261 | ||
1da177e4 LT |
1262 | /* |
1263 | * This routine checks whether a removable media has been changed, | |
1264 | * and invalidates all buffer-cache-entries in that case. This | |
1265 | * is a relatively slow routine, so we have to try to minimize using | |
1266 | * it. Thus it is called only upon a 'mount' or 'open'. This | |
1267 | * is the best way of combining speed and utility, I think. | |
1268 | * People changing diskettes in the middle of an operation deserve | |
1269 | * to lose :-) | |
1270 | */ | |
1271 | int check_disk_change(struct block_device *bdev) | |
1272 | { | |
1273 | struct gendisk *disk = bdev->bd_disk; | |
83d5cde4 | 1274 | const struct block_device_operations *bdops = disk->fops; |
77ea887e | 1275 | unsigned int events; |
1da177e4 | 1276 | |
77ea887e TH |
1277 | events = disk_clear_events(disk, DISK_EVENT_MEDIA_CHANGE | |
1278 | DISK_EVENT_EJECT_REQUEST); | |
1279 | if (!(events & DISK_EVENT_MEDIA_CHANGE)) | |
1da177e4 LT |
1280 | return 0; |
1281 | ||
93b270f7 | 1282 | flush_disk(bdev, true); |
1da177e4 LT |
1283 | if (bdops->revalidate_disk) |
1284 | bdops->revalidate_disk(bdev->bd_disk); | |
1da177e4 LT |
1285 | return 1; |
1286 | } | |
1287 | ||
1288 | EXPORT_SYMBOL(check_disk_change); | |
1289 | ||
1290 | void bd_set_size(struct block_device *bdev, loff_t size) | |
1291 | { | |
e1defc4f | 1292 | unsigned bsize = bdev_logical_block_size(bdev); |
1da177e4 | 1293 | |
5955102c | 1294 | inode_lock(bdev->bd_inode); |
d646a02a | 1295 | i_size_write(bdev->bd_inode, size); |
5955102c | 1296 | inode_unlock(bdev->bd_inode); |
09cbfeaf | 1297 | while (bsize < PAGE_SIZE) { |
1da177e4 LT |
1298 | if (size & bsize) |
1299 | break; | |
1300 | bsize <<= 1; | |
1301 | } | |
1302 | bdev->bd_block_size = bsize; | |
1303 | bdev->bd_inode->i_blkbits = blksize_bits(bsize); | |
1304 | } | |
1305 | EXPORT_SYMBOL(bd_set_size); | |
1306 | ||
4385bab1 | 1307 | static void __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part); |
37be4124 | 1308 | |
6d740cd5 PZ |
1309 | /* |
1310 | * bd_mutex locking: | |
1311 | * | |
1312 | * mutex_lock(part->bd_mutex) | |
1313 | * mutex_lock_nested(whole->bd_mutex, 1) | |
1314 | */ | |
1315 | ||
572c4892 | 1316 | static int __blkdev_get(struct block_device *bdev, fmode_t mode, int for_part) |
1da177e4 | 1317 | { |
1da177e4 | 1318 | struct gendisk *disk; |
523e1d39 | 1319 | struct module *owner; |
7db9cfd3 | 1320 | int ret; |
cf771cb5 | 1321 | int partno; |
fe6e9c1f AV |
1322 | int perm = 0; |
1323 | ||
572c4892 | 1324 | if (mode & FMODE_READ) |
fe6e9c1f | 1325 | perm |= MAY_READ; |
572c4892 | 1326 | if (mode & FMODE_WRITE) |
fe6e9c1f AV |
1327 | perm |= MAY_WRITE; |
1328 | /* | |
1329 | * hooks: /n/, see "layering violations". | |
1330 | */ | |
b7300b78 CW |
1331 | if (!for_part) { |
1332 | ret = devcgroup_inode_permission(bdev->bd_inode, perm); | |
1333 | if (ret != 0) { | |
1334 | bdput(bdev); | |
1335 | return ret; | |
1336 | } | |
82666020 | 1337 | } |
7db9cfd3 | 1338 | |
d3374825 | 1339 | restart: |
0762b8bd | 1340 | |
89f97496 | 1341 | ret = -ENXIO; |
cf771cb5 | 1342 | disk = get_gendisk(bdev->bd_dev, &partno); |
0762b8bd | 1343 | if (!disk) |
6e9624b8 | 1344 | goto out; |
523e1d39 | 1345 | owner = disk->fops->owner; |
1da177e4 | 1346 | |
69e02c59 | 1347 | disk_block_events(disk); |
6796bf54 | 1348 | mutex_lock_nested(&bdev->bd_mutex, for_part); |
1da177e4 LT |
1349 | if (!bdev->bd_openers) { |
1350 | bdev->bd_disk = disk; | |
87192a2a | 1351 | bdev->bd_queue = disk->queue; |
1da177e4 | 1352 | bdev->bd_contains = bdev; |
03cdadb0 | 1353 | |
cf771cb5 | 1354 | if (!partno) { |
89f97496 TH |
1355 | ret = -ENXIO; |
1356 | bdev->bd_part = disk_get_part(disk, partno); | |
1357 | if (!bdev->bd_part) | |
1358 | goto out_clear; | |
1359 | ||
1196f8b8 | 1360 | ret = 0; |
1da177e4 | 1361 | if (disk->fops->open) { |
572c4892 | 1362 | ret = disk->fops->open(bdev, mode); |
d3374825 N |
1363 | if (ret == -ERESTARTSYS) { |
1364 | /* Lost a race with 'disk' being | |
1365 | * deleted, try again. | |
1366 | * See md.c | |
1367 | */ | |
1368 | disk_put_part(bdev->bd_part); | |
1369 | bdev->bd_part = NULL; | |
d3374825 | 1370 | bdev->bd_disk = NULL; |
87192a2a | 1371 | bdev->bd_queue = NULL; |
d3374825 | 1372 | mutex_unlock(&bdev->bd_mutex); |
69e02c59 | 1373 | disk_unblock_events(disk); |
69e02c59 | 1374 | put_disk(disk); |
523e1d39 | 1375 | module_put(owner); |
d3374825 N |
1376 | goto restart; |
1377 | } | |
1da177e4 | 1378 | } |
7e69723f | 1379 | |
22375701 | 1380 | if (!ret) |
7e69723f | 1381 | bd_set_size(bdev,(loff_t)get_capacity(disk)<<9); |
7e69723f | 1382 | |
1196f8b8 TH |
1383 | /* |
1384 | * If the device is invalidated, rescan partition | |
1385 | * if open succeeded or failed with -ENOMEDIUM. | |
1386 | * The latter is necessary to prevent ghost | |
1387 | * partitions on a removed medium. | |
1388 | */ | |
fe316bf2 JN |
1389 | if (bdev->bd_invalidated) { |
1390 | if (!ret) | |
1391 | rescan_partitions(disk, bdev); | |
1392 | else if (ret == -ENOMEDIUM) | |
1393 | invalidate_partitions(disk, bdev); | |
1394 | } | |
5a023cdb | 1395 | |
1196f8b8 TH |
1396 | if (ret) |
1397 | goto out_clear; | |
1da177e4 | 1398 | } else { |
1da177e4 LT |
1399 | struct block_device *whole; |
1400 | whole = bdget_disk(disk, 0); | |
1401 | ret = -ENOMEM; | |
1402 | if (!whole) | |
0762b8bd | 1403 | goto out_clear; |
37be4124 | 1404 | BUG_ON(for_part); |
572c4892 | 1405 | ret = __blkdev_get(whole, mode, 1); |
1da177e4 | 1406 | if (ret) |
0762b8bd | 1407 | goto out_clear; |
1da177e4 | 1408 | bdev->bd_contains = whole; |
89f97496 | 1409 | bdev->bd_part = disk_get_part(disk, partno); |
e71bf0d0 | 1410 | if (!(disk->flags & GENHD_FL_UP) || |
89f97496 | 1411 | !bdev->bd_part || !bdev->bd_part->nr_sects) { |
1da177e4 | 1412 | ret = -ENXIO; |
0762b8bd | 1413 | goto out_clear; |
1da177e4 | 1414 | } |
89f97496 | 1415 | bd_set_size(bdev, (loff_t)bdev->bd_part->nr_sects << 9); |
1da177e4 LT |
1416 | } |
1417 | } else { | |
1da177e4 | 1418 | if (bdev->bd_contains == bdev) { |
1196f8b8 TH |
1419 | ret = 0; |
1420 | if (bdev->bd_disk->fops->open) | |
572c4892 | 1421 | ret = bdev->bd_disk->fops->open(bdev, mode); |
1196f8b8 | 1422 | /* the same as first opener case, read comment there */ |
fe316bf2 JN |
1423 | if (bdev->bd_invalidated) { |
1424 | if (!ret) | |
1425 | rescan_partitions(bdev->bd_disk, bdev); | |
1426 | else if (ret == -ENOMEDIUM) | |
1427 | invalidate_partitions(bdev->bd_disk, bdev); | |
1428 | } | |
1196f8b8 TH |
1429 | if (ret) |
1430 | goto out_unlock_bdev; | |
1da177e4 | 1431 | } |
69e02c59 | 1432 | /* only one opener holds refs to the module and disk */ |
69e02c59 | 1433 | put_disk(disk); |
523e1d39 | 1434 | module_put(owner); |
1da177e4 LT |
1435 | } |
1436 | bdev->bd_openers++; | |
37be4124 N |
1437 | if (for_part) |
1438 | bdev->bd_part_count++; | |
c039e313 | 1439 | mutex_unlock(&bdev->bd_mutex); |
69e02c59 | 1440 | disk_unblock_events(disk); |
1da177e4 LT |
1441 | return 0; |
1442 | ||
0762b8bd | 1443 | out_clear: |
89f97496 | 1444 | disk_put_part(bdev->bd_part); |
1da177e4 | 1445 | bdev->bd_disk = NULL; |
0762b8bd | 1446 | bdev->bd_part = NULL; |
87192a2a | 1447 | bdev->bd_queue = NULL; |
1da177e4 | 1448 | if (bdev != bdev->bd_contains) |
572c4892 | 1449 | __blkdev_put(bdev->bd_contains, mode, 1); |
1da177e4 | 1450 | bdev->bd_contains = NULL; |
0762b8bd | 1451 | out_unlock_bdev: |
c039e313 | 1452 | mutex_unlock(&bdev->bd_mutex); |
69e02c59 | 1453 | disk_unblock_events(disk); |
0762b8bd | 1454 | put_disk(disk); |
523e1d39 | 1455 | module_put(owner); |
4345caba | 1456 | out: |
0762b8bd TH |
1457 | bdput(bdev); |
1458 | ||
1da177e4 LT |
1459 | return ret; |
1460 | } | |
1461 | ||
d4d77629 TH |
1462 | /** |
1463 | * blkdev_get - open a block device | |
1464 | * @bdev: block_device to open | |
1465 | * @mode: FMODE_* mask | |
1466 | * @holder: exclusive holder identifier | |
1467 | * | |
1468 | * Open @bdev with @mode. If @mode includes %FMODE_EXCL, @bdev is | |
1469 | * open with exclusive access. Specifying %FMODE_EXCL with %NULL | |
1470 | * @holder is invalid. Exclusive opens may nest for the same @holder. | |
1471 | * | |
1472 | * On success, the reference count of @bdev is unchanged. On failure, | |
1473 | * @bdev is put. | |
1474 | * | |
1475 | * CONTEXT: | |
1476 | * Might sleep. | |
1477 | * | |
1478 | * RETURNS: | |
1479 | * 0 on success, -errno on failure. | |
1480 | */ | |
e525fd89 | 1481 | int blkdev_get(struct block_device *bdev, fmode_t mode, void *holder) |
1da177e4 | 1482 | { |
e525fd89 TH |
1483 | struct block_device *whole = NULL; |
1484 | int res; | |
1485 | ||
1486 | WARN_ON_ONCE((mode & FMODE_EXCL) && !holder); | |
1487 | ||
1488 | if ((mode & FMODE_EXCL) && holder) { | |
1489 | whole = bd_start_claiming(bdev, holder); | |
1490 | if (IS_ERR(whole)) { | |
1491 | bdput(bdev); | |
1492 | return PTR_ERR(whole); | |
1493 | } | |
1494 | } | |
1495 | ||
1496 | res = __blkdev_get(bdev, mode, 0); | |
1497 | ||
1498 | if (whole) { | |
d4dc210f TH |
1499 | struct gendisk *disk = whole->bd_disk; |
1500 | ||
6a027eff | 1501 | /* finish claiming */ |
77ea887e | 1502 | mutex_lock(&bdev->bd_mutex); |
6a027eff TH |
1503 | spin_lock(&bdev_lock); |
1504 | ||
77ea887e | 1505 | if (!res) { |
6a027eff TH |
1506 | BUG_ON(!bd_may_claim(bdev, whole, holder)); |
1507 | /* | |
1508 | * Note that for a whole device bd_holders | |
1509 | * will be incremented twice, and bd_holder | |
1510 | * will be set to bd_may_claim before being | |
1511 | * set to holder | |
1512 | */ | |
1513 | whole->bd_holders++; | |
1514 | whole->bd_holder = bd_may_claim; | |
1515 | bdev->bd_holders++; | |
1516 | bdev->bd_holder = holder; | |
1517 | } | |
1518 | ||
1519 | /* tell others that we're done */ | |
1520 | BUG_ON(whole->bd_claiming != holder); | |
1521 | whole->bd_claiming = NULL; | |
1522 | wake_up_bit(&whole->bd_claiming, 0); | |
1523 | ||
1524 | spin_unlock(&bdev_lock); | |
77ea887e TH |
1525 | |
1526 | /* | |
d4dc210f TH |
1527 | * Block event polling for write claims if requested. Any |
1528 | * write holder makes the write_holder state stick until | |
1529 | * all are released. This is good enough and tracking | |
1530 | * individual writeable reference is too fragile given the | |
1531 | * way @mode is used in blkdev_get/put(). | |
77ea887e | 1532 | */ |
4c49ff3f TH |
1533 | if (!res && (mode & FMODE_WRITE) && !bdev->bd_write_holder && |
1534 | (disk->flags & GENHD_FL_BLOCK_EVENTS_ON_EXCL_WRITE)) { | |
77ea887e | 1535 | bdev->bd_write_holder = true; |
d4dc210f | 1536 | disk_block_events(disk); |
77ea887e TH |
1537 | } |
1538 | ||
1539 | mutex_unlock(&bdev->bd_mutex); | |
6a027eff | 1540 | bdput(whole); |
e525fd89 TH |
1541 | } |
1542 | ||
1543 | return res; | |
37be4124 | 1544 | } |
1da177e4 LT |
1545 | EXPORT_SYMBOL(blkdev_get); |
1546 | ||
d4d77629 TH |
1547 | /** |
1548 | * blkdev_get_by_path - open a block device by name | |
1549 | * @path: path to the block device to open | |
1550 | * @mode: FMODE_* mask | |
1551 | * @holder: exclusive holder identifier | |
1552 | * | |
1553 | * Open the blockdevice described by the device file at @path. @mode | |
1554 | * and @holder are identical to blkdev_get(). | |
1555 | * | |
1556 | * On success, the returned block_device has reference count of one. | |
1557 | * | |
1558 | * CONTEXT: | |
1559 | * Might sleep. | |
1560 | * | |
1561 | * RETURNS: | |
1562 | * Pointer to block_device on success, ERR_PTR(-errno) on failure. | |
1563 | */ | |
1564 | struct block_device *blkdev_get_by_path(const char *path, fmode_t mode, | |
1565 | void *holder) | |
1566 | { | |
1567 | struct block_device *bdev; | |
1568 | int err; | |
1569 | ||
1570 | bdev = lookup_bdev(path); | |
1571 | if (IS_ERR(bdev)) | |
1572 | return bdev; | |
1573 | ||
1574 | err = blkdev_get(bdev, mode, holder); | |
1575 | if (err) | |
1576 | return ERR_PTR(err); | |
1577 | ||
e51900f7 CE |
1578 | if ((mode & FMODE_WRITE) && bdev_read_only(bdev)) { |
1579 | blkdev_put(bdev, mode); | |
1580 | return ERR_PTR(-EACCES); | |
1581 | } | |
1582 | ||
d4d77629 TH |
1583 | return bdev; |
1584 | } | |
1585 | EXPORT_SYMBOL(blkdev_get_by_path); | |
1586 | ||
1587 | /** | |
1588 | * blkdev_get_by_dev - open a block device by device number | |
1589 | * @dev: device number of block device to open | |
1590 | * @mode: FMODE_* mask | |
1591 | * @holder: exclusive holder identifier | |
1592 | * | |
1593 | * Open the blockdevice described by device number @dev. @mode and | |
1594 | * @holder are identical to blkdev_get(). | |
1595 | * | |
1596 | * Use it ONLY if you really do not have anything better - i.e. when | |
1597 | * you are behind a truly sucky interface and all you are given is a | |
1598 | * device number. _Never_ to be used for internal purposes. If you | |
1599 | * ever need it - reconsider your API. | |
1600 | * | |
1601 | * On success, the returned block_device has reference count of one. | |
1602 | * | |
1603 | * CONTEXT: | |
1604 | * Might sleep. | |
1605 | * | |
1606 | * RETURNS: | |
1607 | * Pointer to block_device on success, ERR_PTR(-errno) on failure. | |
1608 | */ | |
1609 | struct block_device *blkdev_get_by_dev(dev_t dev, fmode_t mode, void *holder) | |
1610 | { | |
1611 | struct block_device *bdev; | |
1612 | int err; | |
1613 | ||
1614 | bdev = bdget(dev); | |
1615 | if (!bdev) | |
1616 | return ERR_PTR(-ENOMEM); | |
1617 | ||
1618 | err = blkdev_get(bdev, mode, holder); | |
1619 | if (err) | |
1620 | return ERR_PTR(err); | |
1621 | ||
1622 | return bdev; | |
1623 | } | |
1624 | EXPORT_SYMBOL(blkdev_get_by_dev); | |
1625 | ||
1da177e4 LT |
1626 | static int blkdev_open(struct inode * inode, struct file * filp) |
1627 | { | |
1628 | struct block_device *bdev; | |
1da177e4 LT |
1629 | |
1630 | /* | |
1631 | * Preserve backwards compatibility and allow large file access | |
1632 | * even if userspace doesn't ask for it explicitly. Some mkfs | |
1633 | * binary needs it. We might want to drop this workaround | |
1634 | * during an unstable branch. | |
1635 | */ | |
1636 | filp->f_flags |= O_LARGEFILE; | |
1637 | ||
572c4892 AV |
1638 | if (filp->f_flags & O_NDELAY) |
1639 | filp->f_mode |= FMODE_NDELAY; | |
1640 | if (filp->f_flags & O_EXCL) | |
1641 | filp->f_mode |= FMODE_EXCL; | |
1642 | if ((filp->f_flags & O_ACCMODE) == 3) | |
1643 | filp->f_mode |= FMODE_WRITE_IOCTL; | |
1644 | ||
1da177e4 | 1645 | bdev = bd_acquire(inode); |
6a2aae06 PE |
1646 | if (bdev == NULL) |
1647 | return -ENOMEM; | |
1da177e4 | 1648 | |
572c4892 AV |
1649 | filp->f_mapping = bdev->bd_inode->i_mapping; |
1650 | ||
e525fd89 | 1651 | return blkdev_get(bdev, filp->f_mode, filp); |
1da177e4 LT |
1652 | } |
1653 | ||
4385bab1 | 1654 | static void __blkdev_put(struct block_device *bdev, fmode_t mode, int for_part) |
2e7b651d | 1655 | { |
2e7b651d | 1656 | struct gendisk *disk = bdev->bd_disk; |
37be4124 | 1657 | struct block_device *victim = NULL; |
2e7b651d | 1658 | |
6796bf54 | 1659 | mutex_lock_nested(&bdev->bd_mutex, for_part); |
37be4124 N |
1660 | if (for_part) |
1661 | bdev->bd_part_count--; | |
1662 | ||
2e7b651d | 1663 | if (!--bdev->bd_openers) { |
6a027eff | 1664 | WARN_ON_ONCE(bdev->bd_holders); |
2e7b651d PZ |
1665 | sync_blockdev(bdev); |
1666 | kill_bdev(bdev); | |
43d1c0eb ID |
1667 | |
1668 | bdev_write_inode(bdev); | |
564f00f6 | 1669 | /* |
43d1c0eb ID |
1670 | * Detaching bdev inode from its wb in __destroy_inode() |
1671 | * is too late: the queue which embeds its bdi (along with | |
1672 | * root wb) can be gone as soon as we put_disk() below. | |
94007751 | 1673 | */ |
43d1c0eb | 1674 | inode_detach_wb(bdev->bd_inode); |
2e7b651d PZ |
1675 | } |
1676 | if (bdev->bd_contains == bdev) { | |
1677 | if (disk->fops->release) | |
db2a144b | 1678 | disk->fops->release(disk, mode); |
2e7b651d PZ |
1679 | } |
1680 | if (!bdev->bd_openers) { | |
1681 | struct module *owner = disk->fops->owner; | |
1682 | ||
0762b8bd TH |
1683 | disk_put_part(bdev->bd_part); |
1684 | bdev->bd_part = NULL; | |
2e7b651d | 1685 | bdev->bd_disk = NULL; |
37be4124 N |
1686 | if (bdev != bdev->bd_contains) |
1687 | victim = bdev->bd_contains; | |
2e7b651d | 1688 | bdev->bd_contains = NULL; |
523e1d39 TH |
1689 | |
1690 | put_disk(disk); | |
1691 | module_put(owner); | |
2e7b651d | 1692 | } |
2e7b651d PZ |
1693 | mutex_unlock(&bdev->bd_mutex); |
1694 | bdput(bdev); | |
37be4124 | 1695 | if (victim) |
9a1c3542 | 1696 | __blkdev_put(victim, mode, 1); |
2e7b651d PZ |
1697 | } |
1698 | ||
4385bab1 | 1699 | void blkdev_put(struct block_device *bdev, fmode_t mode) |
37be4124 | 1700 | { |
85ef06d1 TH |
1701 | mutex_lock(&bdev->bd_mutex); |
1702 | ||
e525fd89 | 1703 | if (mode & FMODE_EXCL) { |
6a027eff TH |
1704 | bool bdev_free; |
1705 | ||
1706 | /* | |
1707 | * Release a claim on the device. The holder fields | |
1708 | * are protected with bdev_lock. bd_mutex is to | |
1709 | * synchronize disk_holder unlinking. | |
1710 | */ | |
6a027eff TH |
1711 | spin_lock(&bdev_lock); |
1712 | ||
1713 | WARN_ON_ONCE(--bdev->bd_holders < 0); | |
1714 | WARN_ON_ONCE(--bdev->bd_contains->bd_holders < 0); | |
1715 | ||
1716 | /* bd_contains might point to self, check in a separate step */ | |
1717 | if ((bdev_free = !bdev->bd_holders)) | |
1718 | bdev->bd_holder = NULL; | |
1719 | if (!bdev->bd_contains->bd_holders) | |
1720 | bdev->bd_contains->bd_holder = NULL; | |
1721 | ||
1722 | spin_unlock(&bdev_lock); | |
1723 | ||
77ea887e TH |
1724 | /* |
1725 | * If this was the last claim, remove holder link and | |
1726 | * unblock evpoll if it was a write holder. | |
1727 | */ | |
85ef06d1 TH |
1728 | if (bdev_free && bdev->bd_write_holder) { |
1729 | disk_unblock_events(bdev->bd_disk); | |
1730 | bdev->bd_write_holder = false; | |
77ea887e | 1731 | } |
6936217c | 1732 | } |
77ea887e | 1733 | |
85ef06d1 TH |
1734 | /* |
1735 | * Trigger event checking and tell drivers to flush MEDIA_CHANGE | |
1736 | * event. This is to ensure detection of media removal commanded | |
1737 | * from userland - e.g. eject(1). | |
1738 | */ | |
1739 | disk_flush_events(bdev->bd_disk, DISK_EVENT_MEDIA_CHANGE); | |
1740 | ||
1741 | mutex_unlock(&bdev->bd_mutex); | |
1742 | ||
4385bab1 | 1743 | __blkdev_put(bdev, mode, 0); |
37be4124 | 1744 | } |
2e7b651d PZ |
1745 | EXPORT_SYMBOL(blkdev_put); |
1746 | ||
1da177e4 LT |
1747 | static int blkdev_close(struct inode * inode, struct file * filp) |
1748 | { | |
4ebb16ca | 1749 | struct block_device *bdev = I_BDEV(bdev_file_inode(filp)); |
4385bab1 AV |
1750 | blkdev_put(bdev, filp->f_mode); |
1751 | return 0; | |
1da177e4 LT |
1752 | } |
1753 | ||
bb93e3a5 | 1754 | static long block_ioctl(struct file *file, unsigned cmd, unsigned long arg) |
1da177e4 | 1755 | { |
4ebb16ca | 1756 | struct block_device *bdev = I_BDEV(bdev_file_inode(file)); |
56b26add | 1757 | fmode_t mode = file->f_mode; |
fd4ce1ac CH |
1758 | |
1759 | /* | |
1760 | * O_NDELAY can be altered using fcntl(.., F_SETFL, ..), so we have | |
1761 | * to updated it before every ioctl. | |
1762 | */ | |
56b26add | 1763 | if (file->f_flags & O_NDELAY) |
fd4ce1ac CH |
1764 | mode |= FMODE_NDELAY; |
1765 | else | |
1766 | mode &= ~FMODE_NDELAY; | |
1767 | ||
56b26add | 1768 | return blkdev_ioctl(bdev, mode, cmd, arg); |
1da177e4 LT |
1769 | } |
1770 | ||
eef99380 CH |
1771 | /* |
1772 | * Write data to the block device. Only intended for the block device itself | |
1773 | * and the raw driver which basically is a fake block device. | |
1774 | * | |
1775 | * Does not take i_mutex for the write and thus is not for general purpose | |
1776 | * use. | |
1777 | */ | |
1456c0a8 | 1778 | ssize_t blkdev_write_iter(struct kiocb *iocb, struct iov_iter *from) |
eef99380 CH |
1779 | { |
1780 | struct file *file = iocb->ki_filp; | |
4ebb16ca | 1781 | struct inode *bd_inode = bdev_file_inode(file); |
7ec7b94a | 1782 | loff_t size = i_size_read(bd_inode); |
53362a05 | 1783 | struct blk_plug plug; |
eef99380 | 1784 | ssize_t ret; |
5f380c7f | 1785 | |
7ec7b94a AV |
1786 | if (bdev_read_only(I_BDEV(bd_inode))) |
1787 | return -EPERM; | |
5f380c7f | 1788 | |
7ec7b94a | 1789 | if (!iov_iter_count(from)) |
5f380c7f AV |
1790 | return 0; |
1791 | ||
7ec7b94a AV |
1792 | if (iocb->ki_pos >= size) |
1793 | return -ENOSPC; | |
1794 | ||
1795 | iov_iter_truncate(from, size - iocb->ki_pos); | |
eef99380 | 1796 | |
53362a05 | 1797 | blk_start_plug(&plug); |
1456c0a8 | 1798 | ret = __generic_file_write_iter(iocb, from); |
e2592217 CH |
1799 | if (ret > 0) |
1800 | ret = generic_write_sync(iocb, ret); | |
53362a05 | 1801 | blk_finish_plug(&plug); |
eef99380 CH |
1802 | return ret; |
1803 | } | |
1456c0a8 | 1804 | EXPORT_SYMBOL_GPL(blkdev_write_iter); |
eef99380 | 1805 | |
b2de525f | 1806 | ssize_t blkdev_read_iter(struct kiocb *iocb, struct iov_iter *to) |
684c9aae LT |
1807 | { |
1808 | struct file *file = iocb->ki_filp; | |
4ebb16ca | 1809 | struct inode *bd_inode = bdev_file_inode(file); |
684c9aae | 1810 | loff_t size = i_size_read(bd_inode); |
a886038b | 1811 | loff_t pos = iocb->ki_pos; |
684c9aae LT |
1812 | |
1813 | if (pos >= size) | |
1814 | return 0; | |
1815 | ||
1816 | size -= pos; | |
a886038b AV |
1817 | iov_iter_truncate(to, size); |
1818 | return generic_file_read_iter(iocb, to); | |
684c9aae | 1819 | } |
b2de525f | 1820 | EXPORT_SYMBOL_GPL(blkdev_read_iter); |
684c9aae | 1821 | |
87d8fe1e TT |
1822 | /* |
1823 | * Try to release a page associated with block device when the system | |
1824 | * is under memory pressure. | |
1825 | */ | |
1826 | static int blkdev_releasepage(struct page *page, gfp_t wait) | |
1827 | { | |
1828 | struct super_block *super = BDEV_I(page->mapping->host)->bdev.bd_super; | |
1829 | ||
1830 | if (super && super->s_op->bdev_try_to_free_page) | |
1831 | return super->s_op->bdev_try_to_free_page(super, page, wait); | |
1832 | ||
1833 | return try_to_free_buffers(page); | |
1834 | } | |
1835 | ||
7f6d5b52 RZ |
1836 | static int blkdev_writepages(struct address_space *mapping, |
1837 | struct writeback_control *wbc) | |
1838 | { | |
1839 | if (dax_mapping(mapping)) { | |
1840 | struct block_device *bdev = I_BDEV(mapping->host); | |
1841 | ||
1842 | return dax_writeback_mapping_range(mapping, bdev, wbc); | |
1843 | } | |
1844 | return generic_writepages(mapping, wbc); | |
1845 | } | |
1846 | ||
4c54ac62 | 1847 | static const struct address_space_operations def_blk_aops = { |
1da177e4 | 1848 | .readpage = blkdev_readpage, |
447f05bb | 1849 | .readpages = blkdev_readpages, |
1da177e4 | 1850 | .writepage = blkdev_writepage, |
6272b5a5 NP |
1851 | .write_begin = blkdev_write_begin, |
1852 | .write_end = blkdev_write_end, | |
7f6d5b52 | 1853 | .writepages = blkdev_writepages, |
87d8fe1e | 1854 | .releasepage = blkdev_releasepage, |
1da177e4 | 1855 | .direct_IO = blkdev_direct_IO, |
b4597226 | 1856 | .is_dirty_writeback = buffer_check_dirty_writeback, |
1da177e4 LT |
1857 | }; |
1858 | ||
25f4c414 DW |
1859 | #define BLKDEV_FALLOC_FL_SUPPORTED \ |
1860 | (FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE | \ | |
1861 | FALLOC_FL_ZERO_RANGE | FALLOC_FL_NO_HIDE_STALE) | |
1862 | ||
1863 | static long blkdev_fallocate(struct file *file, int mode, loff_t start, | |
1864 | loff_t len) | |
1865 | { | |
1866 | struct block_device *bdev = I_BDEV(bdev_file_inode(file)); | |
1867 | struct request_queue *q = bdev_get_queue(bdev); | |
1868 | struct address_space *mapping; | |
1869 | loff_t end = start + len - 1; | |
1870 | loff_t isize; | |
1871 | int error; | |
1872 | ||
1873 | /* Fail if we don't recognize the flags. */ | |
1874 | if (mode & ~BLKDEV_FALLOC_FL_SUPPORTED) | |
1875 | return -EOPNOTSUPP; | |
1876 | ||
1877 | /* Don't go off the end of the device. */ | |
1878 | isize = i_size_read(bdev->bd_inode); | |
1879 | if (start >= isize) | |
1880 | return -EINVAL; | |
1881 | if (end >= isize) { | |
1882 | if (mode & FALLOC_FL_KEEP_SIZE) { | |
1883 | len = isize - start; | |
1884 | end = start + len - 1; | |
1885 | } else | |
1886 | return -EINVAL; | |
1887 | } | |
1888 | ||
1889 | /* | |
1890 | * Don't allow IO that isn't aligned to logical block size. | |
1891 | */ | |
1892 | if ((start | len) & (bdev_logical_block_size(bdev) - 1)) | |
1893 | return -EINVAL; | |
1894 | ||
1895 | /* Invalidate the page cache, including dirty pages. */ | |
1896 | mapping = bdev->bd_inode->i_mapping; | |
1897 | truncate_inode_pages_range(mapping, start, end); | |
1898 | ||
1899 | switch (mode) { | |
1900 | case FALLOC_FL_ZERO_RANGE: | |
1901 | case FALLOC_FL_ZERO_RANGE | FALLOC_FL_KEEP_SIZE: | |
1902 | error = blkdev_issue_zeroout(bdev, start >> 9, len >> 9, | |
1903 | GFP_KERNEL, false); | |
1904 | break; | |
1905 | case FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE: | |
1906 | /* Only punch if the device can do zeroing discard. */ | |
1907 | if (!blk_queue_discard(q) || !q->limits.discard_zeroes_data) | |
1908 | return -EOPNOTSUPP; | |
1909 | error = blkdev_issue_discard(bdev, start >> 9, len >> 9, | |
1910 | GFP_KERNEL, 0); | |
1911 | break; | |
1912 | case FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE | FALLOC_FL_NO_HIDE_STALE: | |
1913 | if (!blk_queue_discard(q)) | |
1914 | return -EOPNOTSUPP; | |
1915 | error = blkdev_issue_discard(bdev, start >> 9, len >> 9, | |
1916 | GFP_KERNEL, 0); | |
1917 | break; | |
1918 | default: | |
1919 | return -EOPNOTSUPP; | |
1920 | } | |
1921 | if (error) | |
1922 | return error; | |
1923 | ||
1924 | /* | |
1925 | * Invalidate again; if someone wandered in and dirtied a page, | |
1926 | * the caller will be given -EBUSY. The third argument is | |
1927 | * inclusive, so the rounding here is safe. | |
1928 | */ | |
1929 | return invalidate_inode_pages2_range(mapping, | |
1930 | start >> PAGE_SHIFT, | |
1931 | end >> PAGE_SHIFT); | |
1932 | } | |
1933 | ||
4b6f5d20 | 1934 | const struct file_operations def_blk_fops = { |
1da177e4 LT |
1935 | .open = blkdev_open, |
1936 | .release = blkdev_close, | |
1937 | .llseek = block_llseek, | |
a886038b | 1938 | .read_iter = blkdev_read_iter, |
1456c0a8 | 1939 | .write_iter = blkdev_write_iter, |
acc93d30 | 1940 | .mmap = generic_file_mmap, |
b1dd3b28 | 1941 | .fsync = blkdev_fsync, |
bb93e3a5 | 1942 | .unlocked_ioctl = block_ioctl, |
1da177e4 LT |
1943 | #ifdef CONFIG_COMPAT |
1944 | .compat_ioctl = compat_blkdev_ioctl, | |
1945 | #endif | |
1e8b3332 | 1946 | .splice_read = generic_file_splice_read, |
8d020765 | 1947 | .splice_write = iter_file_splice_write, |
25f4c414 | 1948 | .fallocate = blkdev_fallocate, |
1da177e4 LT |
1949 | }; |
1950 | ||
1951 | int ioctl_by_bdev(struct block_device *bdev, unsigned cmd, unsigned long arg) | |
1952 | { | |
1953 | int res; | |
1954 | mm_segment_t old_fs = get_fs(); | |
1955 | set_fs(KERNEL_DS); | |
56b26add | 1956 | res = blkdev_ioctl(bdev, 0, cmd, arg); |
1da177e4 LT |
1957 | set_fs(old_fs); |
1958 | return res; | |
1959 | } | |
1960 | ||
1961 | EXPORT_SYMBOL(ioctl_by_bdev); | |
1962 | ||
1963 | /** | |
1964 | * lookup_bdev - lookup a struct block_device by name | |
94e2959e | 1965 | * @pathname: special file representing the block device |
1da177e4 | 1966 | * |
57d1b536 | 1967 | * Get a reference to the blockdevice at @pathname in the current |
1da177e4 LT |
1968 | * namespace if possible and return it. Return ERR_PTR(error) |
1969 | * otherwise. | |
1970 | */ | |
421748ec | 1971 | struct block_device *lookup_bdev(const char *pathname) |
1da177e4 LT |
1972 | { |
1973 | struct block_device *bdev; | |
1974 | struct inode *inode; | |
421748ec | 1975 | struct path path; |
1da177e4 LT |
1976 | int error; |
1977 | ||
421748ec | 1978 | if (!pathname || !*pathname) |
1da177e4 LT |
1979 | return ERR_PTR(-EINVAL); |
1980 | ||
421748ec | 1981 | error = kern_path(pathname, LOOKUP_FOLLOW, &path); |
1da177e4 LT |
1982 | if (error) |
1983 | return ERR_PTR(error); | |
1984 | ||
bb668734 | 1985 | inode = d_backing_inode(path.dentry); |
1da177e4 LT |
1986 | error = -ENOTBLK; |
1987 | if (!S_ISBLK(inode->i_mode)) | |
1988 | goto fail; | |
1989 | error = -EACCES; | |
a2982cc9 | 1990 | if (!may_open_dev(&path)) |
1da177e4 LT |
1991 | goto fail; |
1992 | error = -ENOMEM; | |
1993 | bdev = bd_acquire(inode); | |
1994 | if (!bdev) | |
1995 | goto fail; | |
1996 | out: | |
421748ec | 1997 | path_put(&path); |
1da177e4 LT |
1998 | return bdev; |
1999 | fail: | |
2000 | bdev = ERR_PTR(error); | |
2001 | goto out; | |
2002 | } | |
d5686b44 | 2003 | EXPORT_SYMBOL(lookup_bdev); |
1da177e4 | 2004 | |
93b270f7 | 2005 | int __invalidate_device(struct block_device *bdev, bool kill_dirty) |
b71e8a4c DH |
2006 | { |
2007 | struct super_block *sb = get_super(bdev); | |
2008 | int res = 0; | |
2009 | ||
2010 | if (sb) { | |
2011 | /* | |
2012 | * no need to lock the super, get_super holds the | |
2013 | * read mutex so the filesystem cannot go away | |
2014 | * under us (->put_super runs with the write lock | |
2015 | * hold). | |
2016 | */ | |
2017 | shrink_dcache_sb(sb); | |
93b270f7 | 2018 | res = invalidate_inodes(sb, kill_dirty); |
b71e8a4c DH |
2019 | drop_super(sb); |
2020 | } | |
f98393a6 | 2021 | invalidate_bdev(bdev); |
b71e8a4c DH |
2022 | return res; |
2023 | } | |
2024 | EXPORT_SYMBOL(__invalidate_device); | |
5c0d6b60 JK |
2025 | |
2026 | void iterate_bdevs(void (*func)(struct block_device *, void *), void *arg) | |
2027 | { | |
2028 | struct inode *inode, *old_inode = NULL; | |
2029 | ||
74278da9 | 2030 | spin_lock(&blockdev_superblock->s_inode_list_lock); |
5c0d6b60 JK |
2031 | list_for_each_entry(inode, &blockdev_superblock->s_inodes, i_sb_list) { |
2032 | struct address_space *mapping = inode->i_mapping; | |
2033 | ||
2034 | spin_lock(&inode->i_lock); | |
2035 | if (inode->i_state & (I_FREEING|I_WILL_FREE|I_NEW) || | |
2036 | mapping->nrpages == 0) { | |
2037 | spin_unlock(&inode->i_lock); | |
2038 | continue; | |
2039 | } | |
2040 | __iget(inode); | |
2041 | spin_unlock(&inode->i_lock); | |
74278da9 | 2042 | spin_unlock(&blockdev_superblock->s_inode_list_lock); |
5c0d6b60 JK |
2043 | /* |
2044 | * We hold a reference to 'inode' so it couldn't have been | |
2045 | * removed from s_inodes list while we dropped the | |
74278da9 | 2046 | * s_inode_list_lock We cannot iput the inode now as we can |
5c0d6b60 | 2047 | * be holding the last reference and we cannot iput it under |
74278da9 | 2048 | * s_inode_list_lock. So we keep the reference and iput it |
5c0d6b60 JK |
2049 | * later. |
2050 | */ | |
2051 | iput(old_inode); | |
2052 | old_inode = inode; | |
2053 | ||
2054 | func(I_BDEV(inode), arg); | |
2055 | ||
74278da9 | 2056 | spin_lock(&blockdev_superblock->s_inode_list_lock); |
5c0d6b60 | 2057 | } |
74278da9 | 2058 | spin_unlock(&blockdev_superblock->s_inode_list_lock); |
5c0d6b60 JK |
2059 | iput(old_inode); |
2060 | } |